Some kids appear to learn faster than others. A few years ago, a group of scientists at Carnegie Mellon University decided to study these rapid learners to see what they are doing differently and if their strategies could help the rest of us.
![Website for Mind/Shift](png/mindshift0b39.png)
But as the scientists began their study, they stumbled upon a fundamental problem: they could not find faster learners. After analyzing the learning rates of 7,000 children and adults using instructional software or playing educational games, the researchers could find no evidence that some students were progressing faster than others. All needed practice to learn something new, and they learned about the same amount from each practice attempt. On average, it was taking both high and low achievers about seven to eight practice exercises to learn a new concept, a rather tiny increment of learning that the researchers call a “knowledge component.”
“Students are starting in different places and ending in different places,” said Ken Koedinger, a cognitive psychologist and director of Carnegie Mellon’s LearnLab, where this research was conducted. “But they’re making progress at the same rates.”
Koedinger and his team’s data analysis was published in the Proceedings of the National Academy of Sciences (PNAS), a peer-reviewed journal of the National Academy of Sciences, in March 2023. The study offers the hope that “anyone can learn anything they want” if they get well-designed practice exercises and put some effort into it. Raw talent, like having a “knack for math” or a “gift for language,” isn’t required.
Koedinger and his colleagues wrote that they were initially “surprised” by the “astonishing amount of regularity in students’ learning rate.” The discovery contradicts our everyday experiences. Some students earn As in algebra, an example mentioned in the paper, and they appear to have learned faster than peers who get Cs.
But as the scientists confirmed their numerical results across 27 datasets, they began to understand that we commonly misinterpret prior knowledge for learning. Some kids already know a lot about a subject before a teacher begins a lesson. They may have already had exposure to fractions by making pancakes at home using measuring cups. The fact that they mastered a fractions unit faster than their peers doesn’t mean they learned faster; they had a head start.
Like watching a marathon
Koedinger likens watching children learn to watching a marathon from the finish line. The first people to cross the finish line aren’t necessarily the fastest when there are staggered starts. A runner who finished sooner might have taken five hours, while another runner who finished later might have taken only four hours. You need to know each runner’s start time to measure the pace.
Koedinger and his colleagues measured each student’s baseline achievement and their incremental gains from that initial mark. This would be very difficult to measure in ordinary classrooms, but with educational software, researchers can sort practice exercises by the knowledge components required to do them, see how many problems students get right initially and track how their accuracy improves over time.
In the LearnLab datasets, students typically used software after some initial instruction in their classrooms, such as a lesson by a teacher or a college reading assignment. The software guided students through practice problems and exercises. Initially, students in the same classrooms had wildly different accuracy rates on the same concepts. The top quarter of students were getting 75 percent of the questions correct, while the bottom quarter of students were getting only 55 percent correct. It’s a gigantic 20 percentage point difference in the starting lines.
However, as students progressed through the computerized practice work, there was barely even one percentage point difference in learning rates. The fastest quarter of students improved their accuracy on each concept (or knowledge component) by about 2.6 percentage points after each practice attempt, while the slowest quarter of students improved by about 1.7 percentage points. It took seven to eight attempts for nearly all students to go from 65 percent accuracy, the average starting place, to 80 percent accuracy, which is what the researchers defined as mastery.
The advantage of a head start
The head start for the high achievers matters. Above average students, who begin above 65 percent accuracy take fewer than four practice attempts to hit the 80 percent threshold. Below average students tend to require more than 13 attempts to hit the same 80 percent threshold. That difference – four versus 13 – can make it seem like students are learning at different paces. But they’re not. Each student, whether high or low, is learning about the same amount from each practice attempt. (The researchers didn’t study children with disabilities, and it’s unknown if their learning rates are different.)
The student data that Koedinger studied comes from educational software that is designed to be interactive and gives students multiple attempts to try things, make mistakes, get feedback and try again. Students learn by doing. Some of the feedback was very basic, like an answer key, alerting students if they got the problem right or wrong. But some of the feedback was sophisticated. Intelligent tutoring systems in math provided hints when students got stuck, offered complete explanations and displayed step-by-step examples.
The conclusion that everyone’s learning rate is similar might apply only to well-designed versions of computerized learning. Koedinger thinks students probably learn at different paces in the analog world of paper and pencil, without the same guided practice and feedback. When students are learning more independently, he says, some might be better at checking their own work and seeking guidance.
Struggling students might be getting fewer “opportunities” to learn in the analog world, Koedinger speculated. That doesn’t necessarily mean that schools and parents should be putting low-achieving students on computers more often. Many students quickly lose motivation to learn on screens and need more human interaction.
Memory ability varies
Learning rates were especially steady in math and science – the subjects that most of the educational software in this study focused on. But researchers noticed more divergence in learning rates in the six datasets that involved the teaching of English and other languages. One was a program that taught the use of the article “the,” which can be arbitrary. (Here’s an example: I’m swimming in the Atlantic Ocean today but in Lake Ontario tomorrow. There’s no “the” before lakes.) Another program taught Chinese vocabulary. Both relied on students’ memory and individual memory processing speeds differ. Memory is important in learning math and science too, but Koedinger said students might be able to compensate with other learning strategies, such as pattern recognition, deduction and induction.
To understand that we all learn at a similar rate is one of the best arguments I’ve seen not to give up on ourselves when we’re failing and falling behind our peers. Koedinger hopes it will inspire teachers to change their attitudes about low achievers in their classrooms, and instead think of them as students who haven’t had the same number of practice opportunities and exposure to ideas that other kids have had. With the right exercises and feedback, and a bit of effort, they can learn too. Perhaps it’s time to revise the old saw about how to get to Carnegie Hall. Instead of practice, practice, practice, I’m going to start saying practice, listen to feedback and practice again (repeat seven times).
This story was written by Jill Barshay and produced by The Hechinger Report, a nonprofit, independent news organization focused on inequality and innovation in education. Sign up for the Hechinger newsletter.
“In the LearnLab datasets, students typically used software after some initial instruction in their classrooms, such as a lesson by a teacher or a college reading assignment. The software guided students through practice problems and exercises. Initially, students in the same classrooms had wildly different accuracy rates on the same concepts. The top quarter of students were getting 75 percent of the questions correct, while the bottom quarter of students were getting only 55 percent correct. It’s a gigantic 20 percentage point difference in the starting lines.”
This isn’t the starting line! This is the line after “some initial instruction”, “practice problems”, and “exercises”. So, the “gigantic 20 percentage point difference” includes how much was learned from the initial instruction, practice problems, and exercises. It’s at least partially a difference in learning rate.
To do this experiment properly, you need to pre-test students prior to instruction. (and you have to be quite clever about it, because those “non-existent” students that learn quickly can learn from a pretest).
This is fascinating and compelling and totally contrary to my experience as both teacher and learner. In all of my classes, I have students who pick up material faster and slower; in every area of study I have undertaken, I have seen students both faster and slower than me. While the results of the study may be correct, it is impossible to see or understand the differences that students come in with. A child who has heard a lot of music from a young age might find the tones in Mandarin easier to learn, while a child who draws might find Chinese characters easier to learn. The idea that all learners learn at the same pace is irrelevant given the reality that every students starts at a different place with a unique background.
If at the beginning some students were achieving a score of 75, they may not have the interest to spend any significant amount of time to pick up the other 5%. This is exactly the wrong message to send. We know from research that gifted kids learn faster and need less repetition. Forcing them into more repetitions is what turns them of to education.